cvatkit 0.1.0

A toolkit for reading and processing CVAT annotation files

CVAT Toolkit

A simple Python package for reading CVAT (Computer Vision Annotation Tool) annotation files.

Features

• Read CVAT annotations from XML or ZIP files
• Parse Polygons, Bounding Boxes, and Masks
• Filter by label or image name
• Clean, typed API

Requirements

• Python >= 3.8
• numpy >= 1.20.0

Installation

pip install -e .

Quick Start

from cvatkit import CvatReader

# Load annotations (XML or ZIP)
reader = CvatReader("annotations.xml")

# Print summary
print(reader)

# Get polygons filtered by label
polygons = reader.get_polygons(label="collapse")

# Access polygon data
for polygon in polygons:
    print(f"Image: {polygon.image_name}")
    print(f"Points: {polygon.points}")  # numpy array (N, 2)
    print(f"Attributes: {polygon.attributes}")

Examples

See demo.ipynb for complete examples including:

• Reading from XML and ZIP files
• Working with polygons, boxes, and masks
• Filtering by label and image name
• Using iterators for large datasets
• Getting statistics and summaries

License

MIT License - see LICENSE file for details.